According to a report from TheLec, Samsung Electronics' foundry division has completed the tape-out process for Tesla's next-generation AI5 autonomous driving chip and is preparing for mass production next year. The same TheLec report cites a Samsung Foundry senior engineer who stated on social media that the AI5 chip has taped out and will be manufactured on Samsung's 2nm process at the Taylor, Texas fab, for use in Tesla's latest products.
Tom's Hardware separately reports that James Kim, a principal engineer at Samsung Foundry, wrote in a LinkedIn post: "The Tesla-Samsung AI5 chip has reached tape-out. It is scheduled to be manufactured at the Taylor fab using our latest 2nm process and will soon be integrated into Tesla's newest products. It has been an honor to collaborate with the outstanding engineers at Tesla Palo Alto and Austin over the past several months."
Per TheLec, Park Sang-hoon, head of the customer design support team at Samsung Electronics' foundry business, said at SAFE Forum 2026 that Samsung completed development of a 2nm chip for "a company specializing in automotive and physical AI" in just eight months. He did not identify the customer, but the timeline is cited by TheLec in connection with the Tesla AI5 project.
According to TheLec, Tesla CEO Elon Musk said on X in April that AI5 had completed tape-out; industry observers cited by the outlet believe the tape-out Musk referred to at the time was completed at TSMC. Tom's Hardware adds that Musk demonstrated the first sample of Tesla's AI5 in mid-April and revealed that the processor will be concurrently made both at TSMC and Samsung Foundry.
Tom's Hardware further reports that "AI5 implemented in a TSMC process technology reached taped out several months ahead of AI5 implemented using a Samsung Foundry" — placing the TSMC tape-out around April and the Samsung tape-out reported in July, consistent with the "months apart" framing in both sources.
TheLec reports that industry sources believe Samsung will use its 2nm 2P process or an enhanced 2P+ process for AI5, with the company planning to begin mass production of the 2P+ process in 2027 or 2028. This sits alongside the earlier statement in the same report that overall AI5 mass production at Samsung is planned for next year, suggesting an initial ramp on the current-generation node ahead of a later 2P+ transition.
TheLec reports that Tesla plans to split AI5 production between TSMC and Samsung Electronics. Tom's Hardware attributes the rationale directly to Musk, noting that "Musk expects AI5 to be one of the most produced chip ever, which is why Tesla plans to use two foundries to make it."
Tom's Hardware describes the AI5 processor module Musk demonstrated in April as integrating a relatively compact accelerator die — roughly half a reticle in size, based on Musk's earlier remarks — alongside 12 SK hynix memory packages that appear to be standard GDDR6 or GDDR7 devices.
Assuming the module indeed uses 12 GDDR6 or GDDR7 ICs, Tom's Hardware calculates the processor would feature a 384-bit memory bus, translating into memory bandwidth ranging from 768 GB/s to 1.536 TB/s depending on the memory technology and transfer rates employed. Tom's Hardware also cites a prior claim from Musk that, in certain workloads, AI5 can deliver performance improvements of up to 40X compared to its predecessor.
| Metric | Value | Source |
|---|---|---|
| Samsung 2nm chip development time (unnamed automotive/physical-AI customer) | 8 months | TheLec |
| Samsung Foundry memory bus width | 384-bit | Tom's Hardware |
| Memory bandwidth range (12x GDDR6/GDDR7) | 768 GB/s – 1.536 TB/s | Tom's Hardware |
| Claimed performance gain vs. predecessor (certain workloads) | up to 40X | Tom's Hardware (citing Musk) |
| Samsung 2P+ process mass production target | 2027 or 2028 | TheLec |
According to Tom's Hardware, AI5 is projected to be used in Tesla cars, Tesla robots, and in Tesla's data centers, and Musk expects it to be one of the most produced chips ever — the stated reason Tesla is using two foundries simultaneously.
Taken together, the two sources describe a chip whose two foundry tracks are running on different clocks: the TSMC version reportedly taped out first, around the time of Musk's April demonstration, while the Samsung version taped out later, with TheLec citing Samsung's own claim of an eight-month development cycle for a 2nm automotive/physical-AI chip. Samsung's near-term Taylor fab production is set against a separate, later milestone — 2P+ mass production in 2027 or 2028 — indicating that the chip now taping out and the more advanced process node cited in the same reporting are not necessarily the same manufacturing generation. Musk's stated rationale for splitting output across TSMC and Samsung — anticipated volume large enough to be "one of the most produced chip[s] ever" across cars, robots, and data centers — is the only justification given in the sourced reporting for the dual-foundry strategy.
According to a report by TechNews (2026-07-14), Intel announced a €5 billion (about US$5.7 billion) investment to modernize its Leixlip, Ireland fab for Intel 3-based Xeon 6 and next-generation Xeon processors. The announcement follows a 2024 global 15% headcount cut that included a voluntary severance program at the same site, according to BusinessNext (bnext.com.tw).
According to Inside.com.tw and CNYES, Tesla's next-generation AI5 chip has completed tape-out and will be built on Samsung's 2nm process at the Taylor, Texas fab, with production split between Samsung and TSMC as Samsung's non-memory division reported a roughly KRW 600 billion loss in Q2.
According to a Bloomberg report carried by TechNews, NVIDIA has cut its list of approved Asian AI chip buyers by more than half under a new whitelist system. The Financial Times, cited by Liberty Times, adds that NVIDIA now conducts on-site audits in Singapore, Malaysia and Japan, with the U.S. Commerce Department involved in oversight.